1. Field of the Invention
The present invention relates to imaging, and more specifically, to image quality assurance systems and methodologies for improving the identification of, and speed of access to, image quality suspects in documents.
2. Background Art
On Jun. 6, 2003, the United States House of Representatives passed “The Check Clearing for the 21st Century Act.” Following the passage of the House Bill, the United States Senate enacted on Jun. 27, 2003, S. 1334, “The Check Truncation Act of 2003.” Both of these legislations are referred to informally as “Check 21” legislation. The law aims to improve the check and remittance payment systems by allowing banks, financial institutions, and remittance payment processors (e.g., utility and insurance companies) to exchange checks electronically instead of the current paper-based exchange. President George W. Bush signed the Check 21 legislation into law on Oct. 28, 2003, and when the law goes into effect on Oct. 28, 2004, a printed image of a check or remittance processing stub will have the same legal standing as its original paper counterpart.
Under current law, banking institutions must physically present and return the original paper checks received from customers and other accountholders. Within recent past years, banks, financial institutions, and remittance payment processors (e.g., utility and insurance companies) have been moving away from paper-based financial document processing environments toward image-based processing environments where images of checks and remittance processing stubs are used to perform data entry, correction and balancing operations. The adoption of such image-based document processing has yielded improvements in processing throughput and reduced labor costs.
The Check 21 law allows banks to voluntarily exchange electronic images over networks, legally acknowledging an electronic financial document as the legal equivalent of a paper financial document. In this check truncation environment, only images of the checks would be electronically transmitted between financial institutions and the paper documents would be held or destroyed by the bank of first deposit. Financial institutions desiring to continue processing paper could receive a substitute check, i.e., a paper reproduction created from an electronic image of the original check, for clearing and settlement and ultimate return to the accountholder. This substitute check would have the legal equivalence of the original paper check.
It is expected that the introduction of this law will allow a financial institution's customers to benefit from new products and services, such as online access to their check images. In addition, improved service is anticipated in view of the significantly reduced response time required for account information about checks, and avoidance of negative economic impacts is improved in view of the significantly reduced dependence on air transportation networks, which are vulnerable to expected or unexpected disruptions (such as strikes, weather, natural disaster, terrorist attacks or other types of crises). There is a 12-month implementation period, which began on Oct. 28, 2003, during which banks can roll out imaging technology and set up electronic exchanges with partner institutions.
In view of these changes, and in order to attempt to achieve an acceptable and accepted standard among the aforementioned institutions, the Accredited Standards Committee (ASC) for Financial Services (X9B) has drafted a proposed financial image interchange standard (Specifications for Electronic Exchange of Check and Image Data, DSTU X9.37-2003, currently under revision, hereinafter referred to as “the X9.37 standard”) to support the exchange of check/document imagery between financial institutions. Although this standard has not been finalized or officially adopted, it has been provisionally selected by the Federal Reserve, the central bank of the United States.
However, one issue that arises with check truncation is the new requirement that some level of image quality assurance (IQA) be performed on the check images prior to image interchange between two financial institutions. The goal of IQA would be to provide some level of assurance that the check image renditions being created by the “sending” financial institution are of suitable quality and legibility to support follow-on financial document processing operations by the “receiving” financial institution.
In addressing this issue, the X9.37 standard has included the specification of a “minimum set” of image quality assurance (IQA) “flags” for each image being transmitted or exchanged. The purpose of these IQA flags is to identify questionable, or suspect, document images that might exhibit one or more image quality defects. Currently, the working definition of an “image quality defect” is some defect in the digital image rendition of the check that prohibits its use as a substitute for the original paper document during the payment process. It will be appreciated that such use prohibition may be defined according to the nature of subsequent processing that is to occur. For example, such prohibition might occur if the image does not meet the Check 21 requirement that the substitute check “accurately represents all of the information on the front and back of the original check as of the time the original check was truncated,” or as indicated, infra, if the defect “can fatally compromise recognition of printed or written information by either intelligent character recognition (ICR) systems or by a data entry operator.”
Typical image quality defects may include such items as: 1) Images of checks with folded corners (obscuring some portion of the document writing), 2) images of documents that are excessively skewed, perhaps causing the top portion of the image to be missing (outside the field of view of the image camera), 3) images that exhibit poor contrast or excessive brightness (reducing the legibility of the document), 4) images of two documents that are overlapped causing one document to be partially obscured, and 5) document images whose compressed image file sizes (in kilobytes) are too small or too large. In many cases, these defects in document image quality, albeit occasional, can fatally compromise recognition of printed or written information by either intelligent character recognition (ICR) systems or by a data entry operator.
It is therefore expected that those equipment vendors who supply image-enabled document processing platforms, as well as those who provide image-enabled document processing applications, will need to provide methods and systems to automatically perform image quality assurance for all images generated or created by their document processing platform/solution. It is envisioned that each document image will be tested for the presence of one or more image “defects”, and these defects will be recorded and then stored with the image file/data prior to image interchange between the two banks.
In the past, companies have performed a variety of methods for automatically detecting image quality defects associated with the document image capture process. The automatic detection of image quality defects generally relies on defining a set of image metrics that can be used to infer probable image quality problems in a document image. Typical image metrics to assess image quality may include: image dimensional data (e.g., image height and width), compressed image packet size, image gray level histograms, image skew angle, hardware/software feed check errors, and other transport status/error detection information.
Recently, some companies have demonstrated a variety of methods for automatically detecting image quality defects associated with the document image capture process, and some advances have been made with regard to the automatic detection of image quality defects in document imaging platforms. However, the known image quality flag implementations currently being proposed by the vendors and the interchange standards committees (e.g., IBM and X9.37) involve the appending or embedding of image quality “flag” bits or data records directly into the existing check image file formats. Examples of the image quality flags set forth in the X9.37 standard are set forth in Appendix A. Although this methodology provides a vehicle for storing image quality flags for each document image, it does not provide a method for image-based document processing applications (e.g., check image processing applications) to quickly identify which images are suspected of containing one or more image quality defects.
Using such a method to store image quality flag data for each check image requires that a software application sequentially search/examine the image flag data records stored in each image file for the presence of one or more image quality flags following image capture. FIG. 1 illustrates such a typical IQA implementation scheme. As seen therein, as each document is imaged during the image capture process 100, one or more image rendition files are created at 102 for the front and rear of each document (i.e., a front image file (FIM) 104, a rear image file (RIM) 106, a second front image file (FI2) 108, a second rear image file (RI2) 110). Additionally, an image index file (IDX) 112 is created so that any of the aforementioned individual image files may be accessed by extracting the data stored in the IDX file 112. Finally, an end-of-file (EOF) semaphore file 114 is created to signal the completion of image capture processing for a batch of documents.
At least one image measure metric is determined at IQA metric generation process 118. Image measure metrics may be related to either the document processing and/or imaging hardware or the imaging software. Examples of hardware-related image measure metrics might include document image height and length, document skew, etc. Software-related image measure metrics might include compressed image packet size and front image dimensions compared to rear image dimensions.
Image quality flag (IQF) parameters such as discussed above and as set forth in Appendix A, and predetermined thresholds based on these parameters (120), are applied to image measures determined by the IQA metric generation process 118 at 122 in order to determine whether or not IQA flags (124) are to be set. If IQA flags are to be set, then they are embedded directly in the appropriate image files (i.e., FIM, RIM, FI2, RI2) at 126.
While this methodology may be sufficient for a small number of checks, it is not a particularly efficient method for use with large numbers of checks, particularly in view of the fact that the vast majority of check images do not have any quality defects present. As a result, this methodology results in the wasting of a large amount of application processing time examining image quality flag data where no defects are present. Furthermore, for those check image-processing platforms capable of generating multiple image renditions, e.g., black/white, full gray scale, full resolution, for the front and rear of each check document, the search for imagery containing image quality defects becomes even more complex and time consuming.
Therefore it would be desirable to provide for the realization of automated IQA on image platforms with improved access speed to image quality suspects, and particularly those implementing the X9.37 standard and its progeny.